/* * Copyright (c) 2018, 2026, Red Hat, Inc. All rights reserved. * Copyright Amazon.com Inc. or its affiliates. All Rights Reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it * under the terms of the GNU General Public License version 2 only, as * published by the Free Software Foundation. * * This code is distributed in the hope that it will be useful, but WITHOUT * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License * version 2 for more details (a copy is included in the LICENSE file that * accompanied this code). * * You should have received a copy of the GNU General Public License version * 2 along with this work; if not, write to the Free Software Foundation, * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. * * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA * or visit www.oracle.com if you need additional information or have any * questions. * */ #include "classfile/javaClasses.inline.hpp" #include "gc/shared/barrierSet.hpp" #include "gc/shenandoah/c2/shenandoahBarrierSetC2.hpp" #include "gc/shenandoah/heuristics/shenandoahHeuristics.hpp" #include "gc/shenandoah/shenandoahHeap.hpp" #include "gc/shenandoah/shenandoahRuntime.hpp" #include "gc/shenandoah/shenandoahThreadLocalData.hpp" #include "opto/arraycopynode.hpp" #include "opto/escape.hpp" #include "opto/graphKit.hpp" #include "opto/idealKit.hpp" #include "opto/macro.hpp" #include "opto/narrowptrnode.hpp" #include "opto/output.hpp" #include "opto/rootnode.hpp" #include "opto/runtime.hpp" ShenandoahBarrierSetC2* ShenandoahBarrierSetC2::bsc2() { return reinterpret_cast(BarrierSet::barrier_set()->barrier_set_c2()); } ShenandoahBarrierSetC2State::ShenandoahBarrierSetC2State(Arena* comp_arena) : BarrierSetC2State(comp_arena), _stubs(new (comp_arena) GrowableArray(comp_arena, 8, 0, nullptr)), _trampoline_stubs_count(0), _stubs_start_offset(0), _stubs_current_total_size(0) { } static void set_barrier_data(C2Access& access, bool load, bool store) { if (!access.is_oop()) { return; } DecoratorSet decorators = access.decorators(); bool tightly_coupled = (decorators & C2_TIGHTLY_COUPLED_ALLOC) != 0; bool in_heap = (decorators & IN_HEAP) != 0; bool on_weak = (decorators & ON_WEAK_OOP_REF) != 0; bool on_phantom = (decorators & ON_PHANTOM_OOP_REF) != 0; if (tightly_coupled) { access.set_barrier_data(ShenandoahBitElided); return; } uint8_t barrier_data = 0; if (load) { if (ShenandoahLoadRefBarrier) { if (on_phantom) { barrier_data |= ShenandoahBitPhantom; } else if (on_weak) { barrier_data |= ShenandoahBitWeak; } else { barrier_data |= ShenandoahBitStrong; } } } if (store) { if (ShenandoahSATBBarrier) { barrier_data |= ShenandoahBitKeepAlive; } if (ShenandoahCardBarrier && in_heap) { barrier_data |= ShenandoahBitCardMark; } } if (!in_heap) { barrier_data |= ShenandoahBitNative; } access.set_barrier_data(barrier_data); } Node* ShenandoahBarrierSetC2::load_at_resolved(C2Access& access, const Type* val_type) const { // 1: Non-reference load, no additional barrier is needed if (!access.is_oop()) { return BarrierSetC2::load_at_resolved(access, val_type); } // 2. Set barrier data for load set_barrier_data(access, /* load = */ true, /* store = */ false); // 3. Correction: If we are reading the value of the referent field of // a Reference object, we need to record the referent resurrection. DecoratorSet decorators = access.decorators(); bool on_weak = (decorators & ON_WEAK_OOP_REF) != 0; bool on_phantom = (decorators & ON_PHANTOM_OOP_REF) != 0; bool no_keepalive = (decorators & AS_NO_KEEPALIVE) != 0; bool needs_keepalive = ((on_weak || on_phantom) && !no_keepalive); if (needs_keepalive) { uint8_t barriers = access.barrier_data() | (ShenandoahSATBBarrier ? ShenandoahBitKeepAlive : 0); access.set_barrier_data(barriers); } return BarrierSetC2::load_at_resolved(access, val_type); } Node* ShenandoahBarrierSetC2::store_at_resolved(C2Access& access, C2AccessValue& val) const { // 1: Non-reference store, no additional barrier is needed if (!access.is_oop()) { return BarrierSetC2::store_at_resolved(access, val); } // 2. Set barrier data for store set_barrier_data(access, /* load = */ false, /* store = */ true); // 3. Correction: avoid keep-alive barriers that should not do keep-alive. DecoratorSet decorators = access.decorators(); bool no_keepalive = (decorators & AS_NO_KEEPALIVE) != 0; if (no_keepalive) { access.set_barrier_data(access.barrier_data() & ~ShenandoahBitKeepAlive); } return BarrierSetC2::store_at_resolved(access, val); } Node* ShenandoahBarrierSetC2::atomic_cmpxchg_val_at_resolved(C2AtomicParseAccess& access, Node* expected_val, Node* new_val, const Type* value_type) const { set_barrier_data(access, /* load = */ true, /* store = */ true); return BarrierSetC2::atomic_cmpxchg_val_at_resolved(access, expected_val, new_val, value_type); } Node* ShenandoahBarrierSetC2::atomic_cmpxchg_bool_at_resolved(C2AtomicParseAccess& access, Node* expected_val, Node* new_val, const Type* value_type) const { set_barrier_data(access, /* load = */ true, /* store = */ true); return BarrierSetC2::atomic_cmpxchg_bool_at_resolved(access, expected_val, new_val, value_type); } Node* ShenandoahBarrierSetC2::atomic_xchg_at_resolved(C2AtomicParseAccess& access, Node* val, const Type* value_type) const { set_barrier_data(access, /* load = */ true, /* store = */ true); return BarrierSetC2::atomic_xchg_at_resolved(access, val, value_type); } bool ShenandoahBarrierSetC2::is_Load(int opcode) { switch (opcode) { case Op_LoadN: case Op_LoadP: return true; default: return false; } } bool ShenandoahBarrierSetC2::is_Store(int opcode) { switch (opcode) { case Op_StoreN: case Op_StoreP: return true; default: return false; } } bool ShenandoahBarrierSetC2::is_LoadStore(int opcode) { switch (opcode) { case Op_CompareAndExchangeN: case Op_CompareAndExchangeP: case Op_WeakCompareAndSwapN: case Op_WeakCompareAndSwapP: case Op_CompareAndSwapN: case Op_CompareAndSwapP: case Op_GetAndSetP: case Op_GetAndSetN: return true; default: return false; } } bool ShenandoahBarrierSetC2::can_remove_load_barrier(Node* root) { // Check if all outs feed into nodes that do not expose the oops to the rest // of the runtime system. In this case, we can elide the LRB barrier. We bail // out with false at the first sight of trouble. ResourceMark rm; VectorSet visited; Node_List worklist; worklist.push(root); while (worklist.size() > 0) { Node* n = worklist.pop(); if (visited.test_set(n->_idx)) { continue; } for (DUIterator_Fast imax, i = n->fast_outs(imax); i < imax; i++) { Node* out = n->fast_out(i); switch (out->Opcode()) { case Op_Phi: case Op_EncodeP: case Op_DecodeN: case Op_CastPP: case Op_CheckCastPP: case Op_AddP: { // Transitive node, check if any other outs are doing anything troublesome. worklist.push(out); break; } case Op_LoadRange: { // Array length is the same in all copies. break; } case Op_LoadKlass: { // Klass is the same in all copies. // We would have liked to assert -UCOH, but there are legitimate klass // loads from native Klass* instances, which are also safe under +UCOH. break; } case Op_LoadNKlass: { // Similar to above, but LoadNKlass is only safe without +UCOH. // With +UCOH, it loads from mark word, which clashes with forwarding pointers. if (!UseCompactObjectHeaders) { break; } return false; } case Op_CmpN: { if (out->in(1) == n && out->in(2)->Opcode() == Op_ConN && out->in(2)->get_narrowcon() == 0) { // Null check, no oop is exposed. break; } if (out->in(2) == n && out->in(1)->Opcode() == Op_ConN && out->in(1)->get_narrowcon() == 0) { // Null check, no oop is exposed. break; } return false; } case Op_CmpP: { if (out->in(1) == n && out->in(2)->Opcode() == Op_ConP && out->in(2)->get_ptr() == 0) { // Null check, no oop is exposed. break; } if (out->in(2) == n && out->in(1)->Opcode() == Op_ConP && out->in(1)->get_ptr() == 0) { // Null check, no oop is exposed. break; } return false; } case Op_CallStaticJava: { if (out->as_CallStaticJava()->is_uncommon_trap()) { // Local feeds into uncommon trap. Deopt machinery handles barriers itself. break; } return false; } default: { // Paranoidly distrust any other nodes. return false; } } } } // Nothing troublesome found. return true; } uint8_t ShenandoahBarrierSetC2::refine_load(Node* n, uint8_t bd) { assert(ShenandoahElideIdealBarriers, "Checked by caller"); assert(bd != 0, "Checked by caller"); // Do not touch weak loads at all: they are responsible for shielding from // Reference.referent resurrection. if ((bd & (ShenandoahBitWeak | ShenandoahBitPhantom)) != 0) { return bd; } if (((bd & ShenandoahBitStrong) != 0) && can_remove_load_barrier(n)) { bd &= ~ShenandoahBitStrong; } return bd; } uint8_t ShenandoahBarrierSetC2::refine_store(Node* n, uint8_t bd) { assert(ShenandoahElideIdealBarriers, "Checked by caller"); assert(bd != 0, "Checked by caller"); assert(n->is_Mem() || n->is_LoadStore(), "Sanity"); const Node* newval = n->in(MemNode::ValueIn); assert(newval != nullptr, "Should be present"); // Type system tells us something about nullity? const Type* newval_bottom = newval->bottom_type(); assert(newval_bottom->isa_oopptr() || newval_bottom->isa_narrowoop() || newval_bottom == TypePtr::NULL_PTR, "Should be an oop store"); const TypePtr* newval_type = newval_bottom->make_ptr(); assert(newval_type != nullptr, "Should have been filtered before"); TypePtr::PTR newval_type_ptr = newval_type->ptr(); if (newval_type_ptr == TypePtr::Null) { bd &= ~ShenandoahBitNotNull; // Card table barrier is not needed if we store null. bd &= ~ShenandoahBitCardMark; } else if (newval_type_ptr == TypePtr::NotNull) { // Definitely not null. bd |= ShenandoahBitNotNull; } return bd; } void ShenandoahBarrierSetC2::final_refinement(Compile* compile) const { ResourceMark rm; Unique_Node_List wq; RootNode* root = compile->root(); wq.push(root); // Also seed the outs to capture nodes are not reachable from in()-s, e.g. endless loops. for (DUIterator_Fast imax, i = root->fast_outs(imax); i < imax; i++) { Node* m = root->fast_out(i); wq.push(m); } for (uint next = 0; next < wq.size(); next++) { Node* n = wq.at(next); assert(!n->is_Mach(), "No Mach nodes here yet"); int opc = n->Opcode(); bool is_load = is_Load(opc); bool is_store = is_Store(opc); bool is_load_store = is_LoadStore(opc); uint8_t orig_bd = 0; if (is_load_store) { orig_bd = n->as_LoadStore()->barrier_data(); } else if (is_load || is_store) { orig_bd = n->as_Mem()->barrier_data(); } uint8_t bd = orig_bd; if (ShenandoahElideIdealBarriers && bd != 0) { // Note: we cannot apply load optimizations to LoadStores, // because their load barriers are needed for fixups. if (is_load) { bd = refine_load(n, bd); } if (is_store || is_load_store) { bd = refine_store(n, bd); } } // If there are no real barrier flags on the node, strip away additional fluff. // Matcher does not care about this, and we would like to avoid invoking "barrier_data() != 0" // rules when the only flags are the irrelevant fluff. if ((bd != 0) && (bd & ShenandoahBitsReal) == 0) { bd = 0; } if (bd != orig_bd) { if (is_load_store) { n->as_LoadStore()->set_barrier_data(bd); } else { n->as_Mem()->set_barrier_data(bd); } } for (uint j = 0; j < n->req(); j++) { Node* in = n->in(j); if (in != nullptr) { wq.push(in); } } } } // Support for macro expanded GC barriers void ShenandoahBarrierSetC2::eliminate_gc_barrier_data(Node* node) const { if (node->is_LoadStore()) { LoadStoreNode* loadstore = node->as_LoadStore(); loadstore->set_barrier_data(0); } else if (node->is_Mem()) { MemNode* mem = node->as_Mem(); mem->set_barrier_data(0); } } void ShenandoahBarrierSetC2::eliminate_gc_barrier(PhaseMacroExpand* macro, Node* node) const { eliminate_gc_barrier_data(node); } void ShenandoahBarrierSetC2::elide_dominated_barrier(MachNode* node, MachNode* dominator) const { uint8_t orig_bd = node->barrier_data(); if (orig_bd == 0) { // Nothing to do. return; } uint8_t bd = orig_bd; int node_opcode = node->ideal_Opcode(); if (dominator == nullptr) { // Must be allocation node. if (is_Load(node_opcode) || is_LoadStore(node_opcode)) { // Loads from recent allocations do not need LRBs. bd &= ~ShenandoahBitStrong; } if (is_Store(node_opcode) || is_LoadStore(node_opcode)) { // Stores to recent allocations do not need KA or CM. bd &= ~ShenandoahBitKeepAlive; bd &= ~ShenandoahBitCardMark; } } else { // LoadStores do not get these optimizations, since their LRBs // are required for fixups. if (is_Load(node_opcode) || is_Store(node_opcode)) { int dom_opcode = dominator->ideal_Opcode(); uint8_t dom_bd = dominator->barrier_data(); if (is_Load(dom_opcode) || is_LoadStore(dom_opcode)) { // If dominating load is set up to perform LRB fixups, no further LRB is needed. if ((dom_bd & ShenandoahBitStrong) != 0) { bd &= ~ShenandoahBitStrong; } } if (is_Store(dom_opcode)) { // Dominating store has stored the good ref, no LRB is needed. bd &= ~ShenandoahBitStrong; } } } if (orig_bd != bd) { // We are already in final output. // Strip the extra barrier data if no real bits are left. if ((bd & ShenandoahBitsReal) != 0) { node->set_barrier_data(bd); } else { node->set_barrier_data(0); } } } void ShenandoahBarrierSetC2::analyze_dominating_barriers() const { if (!ShenandoahElideMachBarriers) { return; } ResourceMark rm; Node_List accesses, dominators; PhaseCFG* const cfg = Compile::current()->cfg(); for (uint i = 0; i < cfg->number_of_blocks(); ++i) { const Block* const block = cfg->get_block(i); for (uint j = 0; j < block->number_of_nodes(); ++j) { Node* const node = block->get_node(j); // Everything that happens in allocations does not need barriers. // Record them for dominance analysis. if (node->is_Phi() && is_allocation(node)) { dominators.push(node); continue; } if (!node->is_Mach()) { continue; } MachNode* const mach = node->as_Mach(); int opcode = mach->ideal_Opcode(); if (is_Load(opcode) || is_Store(opcode) || is_LoadStore(opcode)) { if ((mach->barrier_data() & ShenandoahBitsReal) != 0) { accesses.push(mach); dominators.push(mach); } } } } elide_dominated_barriers(accesses, dominators); } uint ShenandoahBarrierSetC2::estimated_barrier_size(const Node* node) const { // Barrier impact on fast-path is driven by GC state checks emitted very late. // These checks are tight load-test-branch sequences, with no impact on C2 graph // size. Limiting unrolling in presence of GC barriers might turn some loops // tighter than with default unrolling, which may benefit performance due to denser // code. Testing shows it is still counter-productive. // Therefore, we report zero barrier size to let C2 do its normal thing. return 0; } bool ShenandoahBarrierSetC2::array_copy_requires_gc_barriers(bool tightly_coupled_alloc, BasicType type, bool is_clone, bool is_clone_instance, ArrayCopyPhase phase) const { bool is_oop = is_reference_type(type); if (!is_oop) { return false; } if (ShenandoahSATBBarrier && tightly_coupled_alloc) { if (phase == Optimization) { return false; } return !is_clone; } return true; } bool ShenandoahBarrierSetC2::clone_needs_barrier(const TypeOopPtr* src_type, bool& is_oop_array) { if (!ShenandoahCloneBarrier) { return false; } if (src_type->isa_instptr() != nullptr) { // Instance: need barrier only if there is a possibility of having an oop anywhere in it. ciInstanceKlass* ik = src_type->is_instptr()->instance_klass(); if ((src_type->klass_is_exact() || !ik->has_subklass()) && !ik->has_injected_fields() && !ik->has_object_fields()) { if (!src_type->klass_is_exact()) { // Class is *currently* the leaf in the hierarchy. // Record the dependency so that we deopt if this does not hold in future. Compile::current()->dependencies()->assert_leaf_type(ik); } return false; } } else if (src_type->isa_aryptr() != nullptr) { // Array: need barrier only if array is oop-bearing. BasicType src_elem = src_type->isa_aryptr()->elem()->array_element_basic_type(); if (is_reference_type(src_elem, true)) { is_oop_array = true; } else { return false; } } // Assume the worst. return true; } void ShenandoahBarrierSetC2::clone(GraphKit* kit, Node* src_base, Node* dst_base, Node* size, bool is_array) const { const TypeOopPtr* src_type = kit->gvn().type(src_base)->is_oopptr(); bool is_oop_array = false; if (!clone_needs_barrier(src_type, is_oop_array)) { // No barrier is needed? Just do what common BarrierSetC2 wants with it. BarrierSetC2::clone(kit, src_base, dst_base, size, is_array); return; } if (ShenandoahCloneRuntime || !is_array || !is_oop_array) { // Looks like an instance? Prepare the instance clone. This would either // be exploded into individual accesses or be left as runtime call. // Common BarrierSetC2 prepares everything for both cases. BarrierSetC2::clone(kit, src_base, dst_base, size, is_array); return; } // We are cloning the oop array. Prepare to call the normal arraycopy stub // after the expansion. Normal stub takes the number of actual type-sized // elements to copy after the base, compute the count here. Node* offset = kit->MakeConX(arrayOopDesc::base_offset_in_bytes(UseCompressedOops ? T_NARROWOOP : T_OBJECT)); size = kit->gvn().transform(new SubXNode(size, offset)); size = kit->gvn().transform(new URShiftXNode(size, kit->intcon(LogBytesPerHeapOop))); ArrayCopyNode* ac = ArrayCopyNode::make(kit, false, src_base, offset, dst_base, offset, size, true, false); ac->set_clone_array(); Node* n = kit->gvn().transform(ac); if (n == ac) { ac->set_adr_type(TypeRawPtr::BOTTOM); kit->set_predefined_output_for_runtime_call(ac, ac->in(TypeFunc::Memory), TypeRawPtr::BOTTOM); } else { kit->set_all_memory(n); } } void ShenandoahBarrierSetC2::clone_at_expansion(PhaseMacroExpand* phase, ArrayCopyNode* ac) const { Node* const ctrl = ac->in(TypeFunc::Control); Node* const mem = ac->in(TypeFunc::Memory); Node* const src = ac->in(ArrayCopyNode::Src); Node* const src_offset = ac->in(ArrayCopyNode::SrcPos); Node* const dest = ac->in(ArrayCopyNode::Dest); Node* const dest_offset = ac->in(ArrayCopyNode::DestPos); Node* length = ac->in(ArrayCopyNode::Length); const TypeOopPtr* src_type = phase->igvn().type(src)->is_oopptr(); bool is_oop_array = false; if (!clone_needs_barrier(src_type, is_oop_array)) { // No barrier is needed? Expand to normal HeapWord-sized arraycopy. BarrierSetC2::clone_at_expansion(phase, ac); return; } if (ShenandoahCloneRuntime || !ac->is_clone_array() || !is_oop_array) { // Still looks like an instance? Likely a large instance or reflective // clone with unknown length. Go to runtime and handle it there. clone_in_runtime(phase, ac, ShenandoahRuntime::clone_addr(), "ShenandoahRuntime::clone"); return; } // We are cloning the oop array. Call into normal oop array copy stubs. // Those stubs would call BarrierSetAssembler to handle GC barriers. // This is the full clone, so offsets should equal each other and be at array base. assert(src_offset == dest_offset, "should be equal"); const jlong offset = src_offset->get_long(); const TypeAryPtr* const ary_ptr = src->get_ptr_type()->isa_aryptr(); BasicType bt = ary_ptr->elem()->array_element_basic_type(); assert(offset == arrayOopDesc::base_offset_in_bytes(bt), "should match"); const char* copyfunc_name = "arraycopy"; const address copyfunc_addr = phase->basictype2arraycopy(T_OBJECT, nullptr, nullptr, true, copyfunc_name, true); Node* const call = phase->make_leaf_call(ctrl, mem, OptoRuntime::fast_arraycopy_Type(), copyfunc_addr, copyfunc_name, TypeRawPtr::BOTTOM, phase->basic_plus_adr(src, src_offset), phase->basic_plus_adr(dest, dest_offset), length, phase->top() ); phase->transform_later(call); phase->igvn().replace_node(ac, call); } void* ShenandoahBarrierSetC2::create_barrier_state(Arena* comp_arena) const { return new(comp_arena) ShenandoahBarrierSetC2State(comp_arena); } ShenandoahBarrierSetC2State* ShenandoahBarrierSetC2::state() const { return reinterpret_cast(Compile::current()->barrier_set_state()); } void ShenandoahBarrierSetC2::print_barrier_data(outputStream* os, uint8_t data) { os->print(" Node barriers: "); if ((data & ShenandoahBitStrong) != 0) { data &= ~ShenandoahBitStrong; os->print("strong "); } if ((data & ShenandoahBitWeak) != 0) { data &= ~ShenandoahBitWeak; os->print("weak "); } if ((data & ShenandoahBitPhantom) != 0) { data &= ~ShenandoahBitPhantom; os->print("phantom "); } if ((data & ShenandoahBitKeepAlive) != 0) { data &= ~ShenandoahBitKeepAlive; os->print("keepalive "); } if ((data & ShenandoahBitCardMark) != 0) { data &= ~ShenandoahBitCardMark; os->print("cardmark "); } if ((data & ShenandoahBitNative) != 0) { data &= ~ShenandoahBitNative; os->print("native "); } if ((data & ShenandoahBitNotNull) != 0) { data &= ~ShenandoahBitNotNull; os->print("not-null "); } if ((data & ShenandoahBitElided) != 0) { data &= ~ShenandoahBitElided; os->print("elided "); } os->cr(); if (data > 0) { fatal("Unknown bit!"); } os->print_cr(" GC configuration: %sLRB %sSATB %sClone %sCard", (ShenandoahLoadRefBarrier ? "+" : "-"), (ShenandoahSATBBarrier ? "+" : "-"), (ShenandoahCloneBarrier ? "+" : "-"), (ShenandoahCardBarrier ? "+" : "-") ); } #ifdef ASSERT void ShenandoahBarrierSetC2::verify_gc_barrier_assert(bool cond, const char* msg, uint8_t bd, Node* n) { if (!cond) { stringStream ss; ss.print_cr("%s", msg); ss.print_cr("-----------------"); print_barrier_data(&ss, bd); ss.print_cr("-----------------"); n->dump_bfs(1, nullptr, "", &ss); report_vm_error(__FILE__, __LINE__, ss.as_string()); } } void ShenandoahBarrierSetC2::verify_gc_barriers(Compile* compile, CompilePhase phase) const { if (!ShenandoahVerifyOptoBarriers) { return; } // Verify depending on the barriers actually enabled, allowing verification in passive mode. // Normally, we have _some_ bits set on all accesses. Optimizations may drop some bits, // but only the last optimization step eliminates all remaining metadata flags. Only then // the access data can be completely blank. bool final_phase = (phase == BeforeCodeGen); bool expect_load_barriers = !final_phase && ShenandoahLoadRefBarrier; bool expect_store_barriers = !final_phase && (ShenandoahSATBBarrier || ShenandoahCardBarrier); bool expect_load_store_barriers = expect_load_barriers || expect_store_barriers; bool expect_some_real = final_phase; Unique_Node_List wq; RootNode* root = compile->root(); wq.push(root); // Also seed the outs to capture nodes are not reachable from in()-s, e.g. endless loops. for (DUIterator_Fast imax, i = root->fast_outs(imax); i < imax; i++) { Node* m = root->fast_out(i); wq.push(m); } for (uint next = 0; next < wq.size(); next++) { Node *n = wq.at(next); assert(!n->is_Mach(), "No Mach nodes here yet"); int opc = n->Opcode(); uint8_t bd = 0; const TypePtr* adr_type = nullptr; if (is_Load(opc)) { bd = n->as_Load()->barrier_data(); adr_type = n->as_Load()->adr_type(); } else if (is_Store(opc)) { bd = n->as_Store()->barrier_data(); adr_type = n->as_Store()->adr_type(); } else if (is_LoadStore(opc)) { bd = n->as_LoadStore()->barrier_data(); adr_type = n->as_LoadStore()->adr_type(); } else if (n->is_Mem()) { bd = MemNode::barrier_data(n); verify_gc_barrier_assert(bd == 0, "Other mem nodes should have no barrier data", bd, n); } bool is_weak = (bd & (ShenandoahBitWeak | ShenandoahBitPhantom)) != 0; bool is_native = (bd & ShenandoahBitNative) != 0; bool is_referent = adr_type != nullptr && adr_type->isa_instptr() && adr_type->is_instptr()->instance_klass()->is_subtype_of(Compile::current()->env()->Reference_klass()) && adr_type->is_instptr()->offset() == java_lang_ref_Reference::referent_offset(); bool is_oop_addr = (adr_type != nullptr) && (adr_type->isa_oopptr() || adr_type->isa_narrowoop()); bool is_raw_addr = (adr_type != nullptr) && (adr_type->isa_rawptr() || adr_type->isa_klassptr()); verify_gc_barrier_assert(!expect_some_real || (bd == 0) || (bd & ShenandoahBitsReal) != 0, "Without real barriers, metadata should be stripped at this point", bd, n); if (is_oop_addr) { if (is_Load(opc)) { verify_gc_barrier_assert(!expect_load_barriers || (bd != 0), "Oop load should have barrier data", bd, n); verify_gc_barrier_assert(!is_weak || is_referent, "Weak load only for Reference.referent", bd, n); } else if (is_Store(opc)) { // Reference.referent stores can be without barriers. verify_gc_barrier_assert(!expect_store_barriers || is_referent || (bd != 0), "Oop store should have barrier data", bd, n); } else if (is_LoadStore(opc)) { verify_gc_barrier_assert(!expect_load_store_barriers || (bd != 0), "Oop load-store should have barrier data", bd, n); } } else if (is_raw_addr) { if (is_native) { if (is_Load(opc)) { verify_gc_barrier_assert(!expect_load_barriers || (bd != 0), "Native oop load should have barrier data", bd, n); } if (is_Store(opc)) { verify_gc_barrier_assert(!expect_store_barriers || (bd != 0), "Native oop store should have barrier data", bd, n); } if (is_LoadStore(opc)) { verify_gc_barrier_assert(!expect_load_store_barriers || (bd != 0), "Native oop load-store should have barrier data", bd, n); } } else { // Some Load/Stores are used for T_ADDRESS and/or raw stores, which are supposed not to have barriers. // Some other Load/Stores are emitted for real oops, but on raw addresses via Unsafe. // The distinction on this level is lost, so we cannot really verify this. } } else { if (is_Load(opc) || is_Store(opc) || is_LoadStore(opc)) { verify_gc_barrier_assert(false, "Unclassified access type", bd, n); } } for (uint j = 0; j < n->req(); j++) { Node* in = n->in(j); if (in != nullptr) { wq.push(in); } } } } #endif static ShenandoahBarrierSetC2State* barrier_set_state() { return reinterpret_cast(Compile::current()->barrier_set_state()); } int ShenandoahBarrierSetC2::estimate_stub_size() const { GrowableArray* const stubs = barrier_set_state()->stubs(); assert(stubs->is_empty(), "Lifecycle: no stubs were yet created"); return 0; } void ShenandoahBarrierSetC2::emit_stubs(CodeBuffer& cb) const { MacroAssembler masm(&cb); PhaseOutput* const output = Compile::current()->output(); assert(masm.offset() <= output->buffer_sizing_data()->_code, "Stubs are assumed to be emitted directly after code and code_size is a hard limit on where it can start"); barrier_set_state()->set_stubs_start_offset(masm.offset()); // Stub generation counts all stubs as skipped for the sake of inlining policy. // This is critical for performance, check it. #ifdef ASSERT int offset_before = masm.offset(); int skipped_before = cb.total_skipped_instructions_size(); #endif GrowableArray* const stubs = barrier_set_state()->stubs(); for (int i = 0; i < stubs->length(); i++) { // Make sure there is enough space in the code buffer if (cb.insts()->maybe_expand_to_ensure_remaining(PhaseOutput::MAX_inst_size) && cb.blob() == nullptr) { ciEnv::current()->record_failure("CodeCache is full"); return; } stubs->at(i)->emit_code(masm); } #ifdef ASSERT int offset_after = masm.offset(); int skipped_after = cb.total_skipped_instructions_size(); assert(offset_after - offset_before == skipped_after - skipped_before, "All stubs are counted as skipped. masm: %d - %d = %d, cb: %d - %d = %d", offset_after, offset_before, offset_after - offset_before, skipped_after, skipped_before, skipped_after - skipped_before); #endif masm.flush(); } void ShenandoahBarrierStubC2::register_stub(ShenandoahBarrierStubC2* stub) { if (!Compile::current()->output()->in_scratch_emit_size()) { barrier_set_state()->stubs()->append(stub); } } ShenandoahBarrierStubC2* ShenandoahBarrierStubC2::create(const MachNode* node, Register obj, Address addr, Register tmp1, Register tmp2, bool narrow, bool do_load) { auto* stub = new (Compile::current()->comp_arena()) ShenandoahBarrierStubC2(node, obj, addr, tmp1, tmp2, narrow, do_load); register_stub(stub); return stub; } void ShenandoahBarrierStubC2::load_post(MacroAssembler* masm, const MachNode* node, Register obj, Address addr, Register tmp1, Register tmp2, bool narrow) { // Load post-barrier: // a. Satisfies the need for LRB for normal loads // b. Passes a weak load through LRB-weak // c. Keep-alives a weak load if (needs_slow_barrier(node)) { ShenandoahBarrierStubC2* const stub = create(node, obj, addr, tmp1, tmp2, narrow, /* do_load = */ false); char check = 0; check |= needs_keep_alive_barrier(node) ? ShenandoahHeap::MARKING : 0; check |= needs_load_ref_barrier(node) ? ShenandoahHeap::HAS_FORWARDED : 0; check |= needs_load_ref_barrier_weak(node) ? ShenandoahHeap::WEAK_ROOTS : 0; stub->enter_if_gc_state(*masm, check); } } void ShenandoahBarrierStubC2::store_pre(MacroAssembler* masm, const MachNode* node, Address addr, Register tmp1, Register tmp2, Register tmp3, bool narrow) { // Store pre-barrier: SATB, keep-alive the current memory value. if (needs_slow_barrier(node)) { assert(!needs_load_ref_barrier(node), "Should not be required for stores"); ShenandoahBarrierStubC2* const stub = create(node, tmp1, addr, tmp2, tmp3, narrow, /* do_load = */ true); stub->enter_if_gc_state(*masm, ShenandoahHeap::MARKING); } } void ShenandoahBarrierStubC2::load_store_pre(MacroAssembler* masm, const MachNode* node, Address addr, Register tmp1, Register tmp2, Register tmp3, bool narrow) { // Load/Store pre-barrier: // a. Avoids false positives from CAS encountering to-space memory values. // b. Satisfies the need for LRB for the CAE result. // c. Records old value for the sake of SATB. // // (a) and (b) are covered because load barrier does memory location fixup. // (c) is covered by KA on the current memory value. if (needs_slow_barrier(node)) { ShenandoahBarrierStubC2* const stub = create(node, tmp1, addr, tmp2, tmp3, narrow, /* do_load = */ true); char check = 0; check |= needs_keep_alive_barrier(node) ? ShenandoahHeap::MARKING : 0; check |= needs_load_ref_barrier(node) ? ShenandoahHeap::HAS_FORWARDED : 0; assert(!needs_load_ref_barrier_weak(node), "Not supported for Load/Stores"); stub->enter_if_gc_state(*masm, check); } } void ShenandoahBarrierStubC2::store_post(MacroAssembler* masm, const MachNode* node, Address addr, Register tmp1, Register tmp2) { if (needs_card_barrier(node)) { cardtable(*masm, addr, tmp1, tmp2); } } void ShenandoahBarrierStubC2::load_store_post(MacroAssembler* masm, const MachNode* node, Address addr, Register tmp1, Register tmp2) { store_post(masm, node, addr, tmp1, tmp2); } bool ShenandoahBarrierStubC2::is_live_register(Register reg) { return preserve_set().member(OptoReg::as_OptoReg(reg->as_VMReg())); } Register ShenandoahBarrierStubC2::select_temp_register(bool& selected_live, Register skip_reg1, Register skip_reg2) { Register tmp = noreg; Register fallback_live = noreg; // Try to select non-live first: for (int i = 0; i < available_gp_registers(); i++) { Register r = as_Register(i); if (r != _obj && r != _addr.base() && r != _addr.index() && r != skip_reg1 && r != skip_reg2 && !is_special_register(r)) { if (!is_live_register(r)) { tmp = r; break; } else if (fallback_live == noreg) { fallback_live = r; } } } // If we could not find a non-live register, select the live fallback: if (tmp == noreg) { tmp = fallback_live; selected_live = true; } else { selected_live = false; } assert(tmp != noreg, "successfully selected"); assert_different_registers(tmp, skip_reg1); assert_different_registers(tmp, skip_reg2); assert_different_registers(tmp, _obj); assert_different_registers(tmp, _addr.base()); assert_different_registers(tmp, _addr.index()); return tmp; } address ShenandoahBarrierStubC2::keepalive_runtime_entry_addr() { if (_narrow) { return CAST_FROM_FN_PTR(address, ShenandoahRuntime::write_barrier_pre_narrow); } else { return CAST_FROM_FN_PTR(address, ShenandoahRuntime::write_barrier_pre); } } address ShenandoahBarrierStubC2::lrb_runtime_entry_addr() { bool is_strong = (_node->barrier_data() & ShenandoahBitStrong) != 0; bool is_weak = (_node->barrier_data() & ShenandoahBitWeak) != 0; bool is_phantom = (_node->barrier_data() & ShenandoahBitPhantom) != 0; if (_narrow) { if (is_strong) { return CAST_FROM_FN_PTR(address, ShenandoahRuntime::load_reference_barrier_strong_narrow_narrow); } else if (is_weak) { return CAST_FROM_FN_PTR(address, ShenandoahRuntime::load_reference_barrier_weak_narrow_narrow); } else if (is_phantom) { return CAST_FROM_FN_PTR(address, ShenandoahRuntime::load_reference_barrier_phantom_narrow_narrow); } } else { if (is_strong) { return CAST_FROM_FN_PTR(address, ShenandoahRuntime::load_reference_barrier_strong); } else if (is_weak) { return CAST_FROM_FN_PTR(address, ShenandoahRuntime::load_reference_barrier_weak); } else if (is_phantom) { return CAST_FROM_FN_PTR(address, ShenandoahRuntime::load_reference_barrier_phantom); } } ShouldNotReachHere(); return nullptr; } bool ShenandoahBarrierSetC2State::needs_liveness_data(const MachNode* mach) const { // Nodes that require slow-path stubs need liveness data. return ShenandoahBarrierStubC2::needs_slow_barrier(mach); } bool ShenandoahBarrierSetC2State::needs_livein_data() const { return true; }